Quick Answer: Ex-USSR genealogy research means tracing ancestors through church books, civil registers, military files and repression records from the former Soviet Union and Russian Empire. Start with the historical jurisdiction where each event happened, check originals before online trees, and log every source. An open-source skill for Codex and Claude Code applies this discipline, keeping each claim traceable and separating proof from guesswork.
Most family-history searches in this region fail for a boring reason: the researcher trusts the first matching name. A tree says your great-grandfather's father was Ivan, a forum post agrees, and the story hardens. Then a scan turns up a different parish, a birth year that's off, and a house number that doesn't match. The ex-USSR Kinship Research skill is built around the opposite habit. Here's how it works and what you can learn from its approach, even if you never install it.
Why evidence-first beats tree-first
Online trees are leads, not proof. The skill ranks sources in a strict order: originals first, then official catalogs and indexes, then published document collections, and trees and forums last. That order matters because most tree errors are copied, not invented. One mistaken entry gets pasted into twenty family sites, and each copy looks like confirmation.
The skill handles this with a rule that surprises most researchers: copies and mirrors of one record share an independence_group, so they never count as independent proof. Two websites showing the same baptism entry are one source, not two. Stop counting duplicates as corroboration and your confidence numbers become honest.
Claims get explainable statuses rather than percentages:
- hypothesis: a possibility with no record behind it yet
- single-source: one record supports it
- corroborated: two or more independent records agree
- strongly-corroborated: independent records agree and nothing conflicts
- contradicted: a reliable record points the other way
- unresolved: the evidence doesn't decide the question
Most people skip the last status. Use it. Saying "unresolved" is more useful than forcing a father's name into your tree.
Start with the jurisdiction, not the surname
A village in Podolia in the 1890s sat in the Russian Empire, under a governorate, a diocese and a parish. Today the same village is in Ukraine, and the records may be catalogued under a different administrative unit. If you search by today's map, you'll miss the right archive. The skill resolves every event to the state, province, district and parish that held it on the event date. Ask yourself: who held the register when this baby was born?
Dates need the same care. Russia used the Julian calendar until 1918. In the nineteenth century, Julian dates ran 12 days behind Gregorian ones. A birth you think is 3 March may be recorded as 19 February, and an exact-date match can fail for no reason except the calendar. Check which calendar a register uses before you conclude two records describe different people.
Names are the second trap. The skill builds name-variant matrices covering pre-reform Russian, Ukrainian, Polish, Romanian, German and Yiddish forms, plus patronymics and maiden names. Ukrainian possessive forms, Romanian fiul lui or n. in patronymic formulas, and Church Slavonic spellings all show up in the same parish. Clerks also swapped letters and forms: Grigorie and Gheorghe may be the same man. Variation isn't noise. It's the clerk's language, and your search should expect it.
Reading unindexed church books
Here's where most guides go wrong: they assume the index is the record. For many ex-USSR parishes, the church book on FamilySearch is only a digital image with no transcription. The project's field notes, drawn from a multi-week case, describe a workable method: find the right film, including the mapping from archive reference to film number, skim the key column, then zoom on candidates.
Logging matters as much as reading. Record the film number, the frame, and which column you skimmed. Without that, you can't reproduce your search, and an unreproducible search can't support a conclusion.
The reading cues that link records are where the real work happens:
- Exact birth dates in a marriage register can match a baptism day for day, which ties a bride to a specific child record.
- House numbers persist for decades in village registers. If the same number appears in two records, that's distinguishing evidence.
- Stillbirths are often listed among births, so a child who never lived may appear in your family count.
Marginal annotations deserve attention too. Later Soviet-era notes on old registers record when a certificate was issued. A note like that is evidence the person or a relative was alive and requesting documents at that time, which is often the only proof of survival for someone who vanished from the indexes.
That said, a shared house number is not identity. Namesakes stay as separate person records until distinguishing evidence appears: parents, godparents, an exact birth date or a military unit. Two men named Petro in one village is the normal state of affairs.
Repression and deportation records
These files are the hardest to read, and the skill treats them neutrally. Published document collections and rehabilitation books group people in ways that can hide individual detail. A "basis" column, such as a 25-year sentence, tells you the legal category used, not the facts of the case.
The most useful counter-intuitive point: a family is linked to a deportation operation, not to the local incident used to justify it. A village incident might be the stated reason in a file, while the actual operation covered whole categories of households. Trace the operation's orders and lists, then match your family to them.
When public sources run out, the skill drafts an archive request rather than sending one. Check the current official procedure before you write, because archive access rules change. A good request names the person, the approximate date and place, the document type and the specific collection, and it states your relationship to the subject.
Military databases and places that no longer exist
Military records have a built-in blind spot. WWII loss databases list the dead, so a soldier who survived won't appear. Absence there tells you little. For Galicia and Bukovina, Austro-Hungarian WWI casualty lists can fill some gaps, but only for the years and units they cover.
The skill also builds "surname nests," clusters of people with the same surname from the same birthplace, which often reveal a family group that single searches miss. Pair this with the question of place. Many hamlets were liquidated, and districts were redrawn. "Born in" a place and "grew up in" it are different facts, and records use both loosely. Keep them separate in your notes.
Some archives sit in occupied or war-damaged areas. Access may be impossible, and that's a real limit to state in your report, not a reason to guess.
Comparison: which records answer which question
| Record type | Typical period | Where it's held | Main limitation |
|---|---|---|---|
| Parish church books | 18th century to 1917 (Orthodox areas) | Diocesan and state archives; FamilySearch films | Often unindexed; images only |
| Civil registration (ZAGS) | From the Soviet civil registry decree of 1917–18 onward | Local civil registry offices, regional archives | Recent records have access limits |
| Military casualty lists | WWI (1914–1918), WWII (1941–1945) | Military archives, national loss databases | Survivors absent; partial coverage |
| Repression and deportation files | 1920s to 1950s | Regional archives, published document collections | Grouped records hide individual detail |
Church books answer "who were the parents?" Civil registers answer "when and where did this person marry or die?" Military lists answer "did this man serve?" Repression files answer "what happened to this family?" Start with the record type that matches your question, not the one that's easiest to search.
What the skill won't do
Trust depends on boundaries. The skill refuses to look up current addresses, phones or accounts of living people, and it minimizes their data in any report. It never solves CAPTCHAs or bypasses paywalls, geo-blocks or rate limits. When a CAPTCHA appears, it hands the task to you, and you sign in yourself. It won't upload your documents or DNA data to third-party services without explicit consent.
The repo's own Responsible use rules also cover sensitive findings like adoption or DNA surprises. Those get reported only as far as the question requires. That restraint is good research practice, not just caution.
Frequently Asked Questions
What is ex-USSR genealogy research?
Ex-USSR genealogy research is tracing family history across the former Soviet Union and Russian Empire using church books, civil registers, military files and repression records. It differs from other genealogy because borders, calendars, languages and record custody changed several times. Researchers resolve each event to the jurisdiction that held the record on that date.
How do I find parents in church records?
Find the parish for the child's birth year and the jurisdiction that held it, then locate the correct FamilySearch film or archive. Skim the birth or baptism column for the surname and check the marriage register for exact dates and house numbers. Log every film number and frame so you can reproduce the search.
Why does the historical jurisdiction matter for my ancestor's village?
Borders, dioceses and archives changed across the 1800s and 1900s. A village in the Russian Empire may now be in a different country, and its records may sit under a different administrative unit. Searching by today's map sends you to the wrong archive and leads to false negatives.
Can an AI agent solve CAPTCHAs on archive sites?
No. The Codex and Claude Code skill hands CAPTCHAs to you, and you sign in to archive sites yourself. This is a deliberate rule, both for legal reasons and because archive access terms vary. The agent can still plan searches, read scans you provide and draft requests for you to send.
Ex-USSR genealogy research rewards patience and paperwork more than luck. This week, pick one ancestor, write down the event, its date and its jurisdiction, then find one original record and log it with its film or archive number. Pass this to someone stuck on a missing parent, and ask them to check the calendar before they check the surname.